On Mon, Nov 18, 2024 at 12:02:44PM +0100, Christophe Leroy wrote: > Le 18/11/2024 à 10:20, Andy Shevchenko a écrit : > > On Sat, Nov 16, 2024 at 11:44:35AM +0100, Christophe Leroy wrote: > > > Le 15/11/2024 à 14:38, Andy Shevchenko a écrit : > > > > Remove legacy-of-mm-gpiochip.h header file, replace of_* functions > > > > and structs with appropriate alternatives. > > > > > > Looks like you don't really have an alternative to > > > of_mm_gpiochip_add_data(), you are replacing one single line by 11 new > > > ones, > > > and that is done twice (once for cpm1_gpiochip_add16(), once for > > > cpm1_gpiochip_add32()).
Note, the overall statistics is +1 LoC only! > > True, but that's the issue that we have of_specific API. If someone propose > > the common API for the agnostic approach, it would be nice, but I am not > > the one. And TBH I do not see the advantage of it right now as almost every > > GPIO driver is using its own labeling schema (*). Note, that this patch also > > fixes a memory leak as a side effect. > > Can you explain that in the commit message instead of saying you use > appropriate alternatives that do not exist ? > > Don't hesitate to mention the memory leak it in the commit message as well. > > > *) the legacy API is mostly used by PPC code, do you want that of_mm_* > > thingy > > to be moved to PPC specific code instead of killing it? Would be done this > > way as well. > > No no, your change is ok for me, just need an accurate commit message. Fine, just sent a v2, which is now compile-tested as suggested below. ... > > > Does not build: > > > > Crap, I most likely built something else and not these files... > > I have carefully build-test this for v2. > > Just use mpc885_ads_defconfig Thanks, that is helpful! -- With Best Regards, Andy Shevchenko
